You are on page 1of 1

Title of the Project : Quality Analysis of oranges based on external

Name:- (all students in the batch ) features using convolutional neural network
B-Tech in Electronics &Communication Engineering
Guide :-
College of Engineering, Chengannur, June 2020
Sensory characteristics of fruits and vegetables in appearance is the one that impacts their market value, and consumer’s preference. The project is to implement a stand-alone quality testing system of oranges with pi-4
ABSTRACT: using CNN , to replace the manual inspection system and makes the sorting more accurate,efficient and less time consuming.
INTRODUCTION DESIGN FLOW EXPERIMENTAL RESULTS

Orange is well known for its anti-microbial, anti-viral and other properties. High-quality products are the IMAGE ACQUISITION One of the most promising applications in computer vision is the recognition of fruits using Deep Convolutional
basis for the success in today’s highly competitive market. Currently, manual inspection is being used in Neural Network. In recent times, deep learning based classifications are making it possible to recognize fruits from
order to determine the orange quality. Quality assurance requires simple and reliable sorting methods.The RGB images. However, fruit recognition is still a problem for the stacked fruits on weighing scale because of the
use of computer vision systems enables the detection of external quality defects.This project is to
complexity and similarity. Here deep learning techniques for the classification is used. We have used Fruits-300
investigate the applicability of a method for detection of external defects and an automatic orange’s GRAY IMAGE
classification system based on textural features of images.The system is very robust and can identify dataset for the evaluation purpose. From the dataset, we have established a dataset which contains 300 images
specific defects like:morphological defect, surface defect ,colour defect ,black mould or to recognize a from different categories. The images are divided into training and test dataset. Moreover, for the classification
BINARY IMAGE
normal fruit.
accuracies, we have used various combinations of hidden layer and epochs for different cases and made a
AIM & OBJECTIVE FEATURES EXTRACTION comparison between them. The overall performance losses of the network for different cases also observed.
Finally, we have achieved the best test accuracy of 100% and a training accuracy of 99.79%.
To investigate the applicability of a method for detection of external defects and an automatic orange’s TRAINING
classification system. For this purpose, the development of a classifier based on textural features of TESTING FEATURE FEATURE
TRAIN NN
images, captured with a digital camera, was evaluated. The proposed technique is very robust and can CONCLUSION FUTURE WORKS
identify specific defects like: surface defect, morphological defect, color defect, black mould or recognize
CLASSIFICATION
a normal fruit. Plenty of research carried out for grading citrus fruits using machine vision. Citrus grading
is normally achieved based on external visible criteria Including size, shape, and color of the fruits. BLOCK DIAGRAM System uses universal filter at pre-processing stage. This stage is crucial and is used to enhance the image along
with its clarity. The clarity of the image determines the final result accuracy. Next phase is feature extraction.
LITERATURE REVIEW CONTROL MODULE-PI 4 Features are extracted in the form of textured and colored features. Segmentation through the k-means clustering is
CAMERA LAPTOP/PC
TITLE METHOD ADVANTAGES REMARKS(SPECIFY performed. The primary objective of this approach is to give critical and non-critical image segment division.
PROPOSED AND OPEN ENDS/FUTURE Classification through CNN is obtained. Convolution neural network is yield optimal result in terms of
DISADVANTA WORKS)
classification accuracy.
GES
CONVEYOR ASSEMBLY SORTING ASSEMBLY In future, real time datasets can be tested against the proposed system. Also convergence rate can be further
improved using prompted random forest mechanism.

Shiv.R.D et al(2018)”Application Training-on-site,SID Only a single type of To improve the classification IMPLEMENTATION
of image processing in fruit and classification,NIR-H fruit and disease accuracy several other features
vegetable analysis” SI,SVM considered,average can be included REFERENCES
classification error ● Capture image of the fruit with picam.
was less
● Resize the image.
● Convert the image to grayscale.
H.Cheng et al(2017),”Early yield Image BPNN can be used Separate a site-specific model ● Identify the category of the orange using the trained model. Erdenee, B., Ryutaro, T., Tana, G., 2010, Particular Agricultural Land Cover Classification Case Study Of
prediction using image analysis analysis,ANN,BPN for apple yield from a general model ,investigate
..NN” N, precision prediction accurately the adaptation of the proposed ● Print the category of the orange in the image Tsagaannuur, Mongolia. In: IEEE International Geoscience & Remote Sensing Symposium, 3194-3197.
horticulture model to other tree forms etc.
Tewari, V.K., Arudra, A.K., Kumar, S.P., Pandey, V., Chandel, N.S., 2013. Estimation of plant nitrogen content
TOOLS REQUIRED
using digital image processing. Int. Commission Agricu. Biosyst. Eng. 15 (2), 78–86.
H.M Zawbaa et Pre-processing, Classification Use other feature extraction and Krishna, M., Jabert, G., 2013. Pest control in agriculture plantation using image processing. IOSR J. Electron.
al(2014)“Automatic fruit image feature extraction, accuracy depends on classifiers algorithms in order to SOFTWARE
recognition system based on and classification the differentiations increase the obtained accuracy. Commun. Eng. (IOSR-JECE) 6 (4), 68–74.
shape and color features” phases,SVM,K-NN, degree between the
Tensorflow library- Patil, J.K., Kumar, R., 2011. Advances in image processing for detection of plant diseases. J. Adv. Bioinf. Appl.
SIFT fruit types
Res. ISSN 2 (2), 135–141.
TensorFlow (TF) is an open source API developed by Google mainly for Machine Learning and
Naik, S., Patel, B., 2017. A machine vision based fruit classification and grading: a review. Int. J. Comput. Appl.
S.Lawrence (1997)“Face Hybrid Neural Rapid classification, More careful selection of the Deep Learning, but it is also applicable for other numerical computations.
recognition:A CNN approach” Network,local image no explicit 3-D convolutional network 170 (9), 22–34.
mapping,SOM models architecture, addition of
pseudo-data

You might also like